Galois Cohomology and Algebraic K-theory of Finite Fields
by
Gabriel J. Angelini-Knoll
AN ESSAY
Submitted to the College of Liberal Arts and Sciences,
Wayne State University,
Detroit, Michigan,
in partial fulfillment of the requirements
for the degree of
MASTER OF ARTS
April of 2013
MAJOR: Mathematics
APPROVED BY:Advisor Date
2nd Reader Date(if necessary)
Acknowledgements
I would like to extend my utmost graditude to all the people who helped me with
this project. First, I would like to thank my family for supporting me in any endeavor I
undertake. Second, I would like to thank Jessica McInchak for her patience with me during
tough times and for reading drafts along the way. I am in greatest debt to my advisor
Dr. Andrew Salch who spent countless hours discussing the material presented here and
much more. His enthusiasm for mathematics is an inspiration for me and without him this
would not have been possible. I would like to thank my colleagues in the Mathematics
department especially Sean Tilson, Michael Catanzaro, and Bogdan Gheorghe for advice
on mathematics research and for helping me form relationships with faculty members.
Finally, I would like to thank my friends outside of the mathematics department for helping
me get through the long days of graduate student life.
1
Contents
1 Introduction 4
2 Basic Galois Cohomology 6
2.1 Basic Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Basic Galois Theory . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 G-Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Computing Group Cohomology . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Finite Galois Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Continuous Galois Cohomology 14
3.1 Profinite Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Profinite Group Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Computations in Continuous Galois Cohomology . . . . . . . . . . . . . . 16
4 Algebraic K-theory of finite fields 23
4.1 Continuous Galois Cohomology to Etale Cohomology . . . . . . . . . . . 24
4.1.1 Definitions: basic algebraic geometry . . . . . . . . . . . . . . . . 24
4.1.2 Definitions needed for etale cohomology . . . . . . . . . . . . . . 25
4.1.3 Equivalence of categories and cohomology theories. . . . . . . . . 27
4.2 Norm Residue Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2
4.3 Algebraic K-theory Spectral Sequence . . . . . . . . . . . . . . . . . . . . 32
4.4 Spectral Sequence Computation . . . . . . . . . . . . . . . . . . . . . . . 33
4.5 Comparison to results of Quillen . . . . . . . . . . . . . . . . . . . . . . . 37
5 Future Research 41
3
Chapter 1
Introduction
The motivation for this work comes from recent progress in showing a relationship
between etale cohomology and algebraic K-theory largely due to Voevodsky and Rost, and
in part due to Weibel. This progress begins with the proof of the Milnor conjecture by
Voevodsky in 1996 [15]. Milnor conjectured that for k a field with char(k) , 2, there are
isomorphisms from Milnor K-theory of k modulo 2 to etale cohomology of k with coeffi-
cients in Z/2Z for n ≥ 0 [15]. This suggests that Milnor K-theory can be computed using
etale cohomology in this special case.
A generalization of this conjecture for primes not equal to 2 is the Bloch-Kato con-
jecture (or motivic Bloch-Kato conjecture), which has now been proven in a series of 15-20
papers by Voevodsky, Rost, and Weibel [16]. The proof of this conjecture was announced
at a conference honoring Grothendieck in 2009 and it is now referred to as the norm-residue
theorem. The Bloch-Kato conjecture suggests a relationship between etale cohomology and
algebraic K-theory as well and with the proof of the norm-residue theorem this relationship
was solidified. Thus, for our purposes, it shows that algebraic K-theory is computable via
etale cohomology .
A useful tool we apply is a spectral sequence which has been shown to converge to
4
algebraic K-theory. This story goes back further to 1972 when Lichtenbaum conjectured
about a relationship of K-theory to etale cohomology for the purposes of studying zeta
functions [16]. Later Quillen and Beilinson expanded and generalized these conjectures.
One construction of this spectral sequence came from Friedlander and Suslin in a preprint
in 1994. The most useful versions of the spectral sequence for our purposes can be found
in Weibel’s book on algebraic K-theory, yet to arrive in print [16], and Thomason’s paper
on algebraic K-theory and etale cohomology [13].
All theorems used directly in this paper will be stated in their original generality
and then translated to the special case of finite fields. The goal of this paper is to approach
algebraic K-theory from a different direction than direct computation using some recent re-
sults. We begin with some results from basic galois theory and group cohomology making
specific computations where they will be helpful for later use. Next we define and compute
profinite group cohomology and continuous galois cohomology. We follow by identify-
ing continuous galois cohomology with etale cohomology and then etale cohomology with
motivic cohomology in this special case. We use this information to compute algebraic
K-theory of finite fields which we find to be the following for i ≥ 1
K0(Fp;Z/`Z)[β−1] � Z/`ZK2i(Fp;Z/`Z) � Z/`Z if `|pi − 1.
K2i−1(Fp;Z/`Z) � Z/`Z if `|pi − 1.K∗(Fp;Z/`Z) � 0 otherwise
We show that comparison with Quillen’s original results confirms our answer. We
then suggest other generalizations of this approach.
5
Chapter 2
Basic Galois Cohomology
Since galois cohomology will be useful for our calculations, we begin with a brief
summary of important facts from galois theory and group cohomology and use them to
compute cohomology of finite galois groups. Though we consider the finite case here,
ultimately we will be interested in the profinite case called continuous galois cohomology.
2.1 Basic Facts
We summarize some facts about galois groups and G-modules.
2.1.1 Basic Galois Theory
Let Fpn denote the field of order pn, for p a prime and n ≥ 1 an integer. Recall that any
finite field is isomorphic to Fpn for some prime and some integer n ≥ 1.
Proposition 2.1.1. If Fpn is a finite extension of a finite field Fp, then Fpn is finite and is
galois over K. The galois group Gal(Fpn/Fp) = Gn is cyclic and Gn � Z/nZ.
Proof. See [3] or any other basic abstract algebra text. �
6
Proposition 2.1.2. Let Gmn = Gal(Fpmn/Fp), and Gn = Gal(Fpn/Fp) and let H = Gal(Fpmn/Fpn).
Then
Gn � Gmn/H
Proof. This follows from the fundamental theorem of galois theory; see [3]. �
These facts will be usefull for future calculations. We note here that Prop. 2.1.1
gives us the galois groups as cyclic groups, so it will suffice to understand cohomology of
cyclic groups.
2.1.2 G-Modules
We will be dealing with G-modules where G is the galois group of an extension of Fp.
Definition 2.1.3. Let G be a group. Then a G-module M is an abelian group along with a
(left) action of G, by which we mean a map G×M → M such that the following properties
hold for a, b ∈ M, s, t ∈ G
1. 1.a=a
2. s.(a+b)=s.a+s.b
3. s.(t.a)=(s.t).a
where s.a denotes s acting on a; i.e the map above sends (s,a) to s.a.
2.2 Computing Group Cohomology
Let A be a G-module, then we let AG denote the elements of A fixed by the action
of G. Given a function f : A → B there is an induced map f : AG → BG making AG
functorial. It is an additive left exact functor. We define cohomology in the following way:
7
Definition 2.2.1. Given A a G-module, we take the right derived functors Ri of the functor
AG and define the following
Hi(G, A) = Ri(AG).
We call this cohomology of G with coefficients in A.
We consider cohomology of cyclic groups here as they are the most pertinent for
our goal. Our resources for this material are [1], [4] along with [10] and [11] for the
applications.
Proposition 2.2.2. Let Hi(Z/nZ, A) be group cohomology of Z/nZ with with coefficients in
A. We let A have the trivial action. For A=Z/`Z we get
Hi(Z/nZ,Z/`Z) �
Z/`Z if i = 0
Z/ gcd(n, `)Z if i > 0.
Proof. We let Z[Z/nZ] be the group ring and we use the fact that
Hi(Z/nZ,Z/`Z) � ExtiZ[Z/nZ](Z,Z/`Z).
We then compute Ext in this case. First, we take a projective resolution of Z (suffices in
most cases to consider a free resolution since a free module is a projective module). We
write a general element of the group ring Z[Z/nZ] as a(b) where a ∈ Z and b ∈ Z/nZ.
0←− Zε←− Z[Z/nZ]
D←− Z[Z/nZ]
N←− Z[Z/nZ]←− ...
by letting ε be the the map which sends 1(m) to 1 for all 0 ≤ m ≤ n − 1 (called the
augmentation map). The map D sends 1(0) to 1(1)-1(0). We let N be the map which sends
1(1) to∑n−1
i=0 1(i). The following maps are D again and N and this pattern repeats. We then
8
truncate and apply Hom(−,Z/`Z). Resulting in the sequence,
0 −→ Hom(Z[Z/nZ],Z/`Z)D∗−→ Hom(Z[Z/nZ],Z/`Z)
N∗−→ Hom(Z[Z/nZ],Z/`Z)...
and using a simple fact from algebra HomZ[Z/nZ](Z[Z/nZ],Z/`Z) � Z/`Z [3], we get the
sequence
0 −→ Z/`ZD∗−→ Z/`Z
N∗−→ Z/`Z
D∗−→ Z/`Z
N∗−→ Z/`Z...
We are then left to determine the maps between these modules [ D∗ and N∗]. The isomor-
phism from Hom to Z/`Z takes f ∈ HomZ[Z/nZ](Z[Z/nZ],Z/`Z) to f(1(0)), so the map D*
is ( f ◦D)(1(0)) = f (1(1)− f (1(0)). Since f is a linear Z[Z/nZ] module map and since Z/nZ
acts trivially on Z/`Z, we have
f (1(1) − 1(0)) = f (1(1)) − f (1(0)) = 1(1) f (1(0)) − f (1(0)) = f (1(0)) − f (1(0)) = 0.
Thus, D∗ is the 0 map. We then find N∗. Similarly we have f (1(0)) goes to ( f ◦ N)(1(0))
and for the same reasons
( f ◦ N)(1(0)) = f (1(0) + 1(1) + ... + 1(n − 1))
= f (1(0) + 1(1) f (1(0)) + 1(2) f (1(0) + ... + 1(n − 1) f (1(0))
= n f (1(0)).
Thus, the map N∗ is multiplication by n.
We are now in a position to compute the functor Ext:
ExtiZ[Z/nZ](Z,Z/`Z) � Hi(Z/nZ,Z/`Z)
9
Ext0Z[Z/nZ](Z,Z/`Z) � kerD∗/{0} = Z/`Z
Ext1Z[Z/nZ](Z,Z/`Z) � kerN∗/imD∗ = (Z/ gcd(n, `)Z)/{0} = Z/ gcd(n, l)Z
Ext2Z[Z/nZ](Z,Z/`Z) � kerD∗/imN∗ = (Z/`Z)/nZ = Z/ gcd(n, l)Z
ExtiZ[Z/nZ](Z,Z/`Z) � Z/ gcd(n, `)Z for i ≥ 3 .
This gives us a simple expression in terms of n and l of Hi(Z/nZ,Z/`Z). We have therefore
computed the following:
H0(Z/nZ,Z/`Z) � Z/`Z
Hi(Z/nZ,Z/`Z) � Z/ gcd(n, `) for i> 0.
�
It will also be useful to know cohomology of cyclic groups with coefficients A in
more generality. If g is a map from A to A, then we denote gA the kernel of g contained in
A and gA the image of the map g in A. We state what they are explicitly in the proposition
below.
Proposition 2.2.3. Let G be a cyclic group of order n with generator t and A a G-module
then
Hi(G, A) �
AG/N∗A if i is even
N∗A/D∗A if i is odd
for i > 0. And we have,
H0(G, A) � AG
where
AG = {a ∈ A | t(a) = a for all t ∈ G}N∗A = {a ∈ A | (1 + t + t2 + ... + tn−1)(a′) = a for some a′ ∈ A}N∗A = {a ∈ A | (1 + t + t2 + ... + tn−1)(a) = 0}D∗A = {a ∈ A | (t − 1)(a′) = a for some a′ ∈ A}
10
with t(a) the action of t on a.
Proof. Similar to the previous proof, we use the fact that Hi(G, A) � ExtiZ[G](Z, A). We take
a projective resolution of Z.
0←− Zε←− Z[Z/nZ]
D←− Z[Z/nZ]
N←− Z[Z/nZ]←− ...
where ε, N, and D are the same maps as before. The only difference here is the next step.
We truncate and apply Hom(−, A) to get
0 // Hom(Z[G], A) D∗ //
���
Hom(Z[G], A) N∗ //
���
Hom(Z[G], A)
���
D∗ // Hom(Z[G], A)
���
0 // A D∗ // A N∗ // A D∗ // A
We then need to find the maps D∗ and N∗. We can compute that the map D∗ : A→ A sends
a to a−ta and N∗ : A→ A sends a to a+ta+...+tn−1a. Thus, we have enough to get a general
formula for cohomology. Note that kerD∗ � {a ∈ A | ta − a = 0} � {a ∈ A | ta = a} � AG
H0(G, A) � AG
Hi(G, A) � N∗A/D∗A if i = 2m − 1 m ≥ 1
Hi(G, A) � AG/N∗A if i = 2m m ≥ 1
thus we have shown our claim.
�
2.3 Finite Galois Cohomology
Once we have cohomology of finite cyclic groups, we get the finite galois cohomology we
need by Prop. 2.1.1. The difference that may arise is the action that the galois group has on
11
our coefficient module. We will discuss this more in the next section.
Proposition 2.3.1. Let Gal(Fpn/Fp)=Gn be the galois group of order n and µ`(Fpn) be the
`-th roots of unity in Fpn and suppose `|pn − 1 then we compute galois cohomology as
follows.
Hi(Gn; µl(Fpn)) �
Z/`Z if i = 0
Z/ gcd(n, `)Z if i > 0.
If ` does not divide pn − 1 then µ`(Fpn) = 0.
Proof. The proof follows from Prop. 2.1.1 and Prop. 2.2.2. �
In the above proposition, we use that the order of µ`(Fpn) is gcd(`, pn − 1), which is
a well known fact. We also get a general formula for finite galois cohomology.
Proposition 2.3.2. Let Gn be defined as above, A a Gn-module, then
Hi(Gn, A) �
AGn/N∗A if i is even
N∗A/D∗A if i is odd
for i > 0. And we have,
H0(Gn, A) = AGn
where
AGn = {a ∈ A | g(a) = a for all g ∈ G}N∗A = {a ∈ A | (1 + g + g2 + ... + gn−1)(a) = a}N∗A = {a ∈ A | (1 + g + g2 + ... + gn−1)(a) = 0}D∗A = {a ∈ A | (1 − g)(a) = a}
with g(a) the action of g on a, and g the generator of Gn.
Proof. Combine Prop. 2.1.1 with Prop. 2.2.3. �
12
We use these computations to introduce cohomology and give the flavor of our fu-
ture computations, though we ultimately consider profinite galois cohomology, also called
continuous galois cohomology.
13
Chapter 3
Continuous Galois Cohomology
Here we will introduce continuous galois cohomology, which is cohomology of a
profinite galois group. We lead up to a specific computation that will be useful in the next
section. First, we define profinite groups, then we define cohomology of a profinite galois
group. We then move to specific computations.
3.1 Profinite Groups
For this section our main reference is Serre’s book on galois cohomology [11].
Definition 3.1.1. A profinite group is a topological group defined as the projective limit of
finite groups each with the discrete topology.
We can say more about the topology of a profinite group as well:
Proposition 3.1.2. A profinite group is compact and totally disconnected and a compact
and totally disconnected topological group is a profinite group.
Proof. The first statement follows from the definition and for the converse see [11] �
The important example for us is the following:
14
Example 3.1.3. Let L be a finite galois extension of K, K a field. Then Gal(L/K) is a
finite group. We give Gal(L/K) the discrete topology. We then take the projective limit
lim←−
Gal(L/K) where the limit is taken over all finite extensions L/K. This projective limit is
by definition the galois group Gal(Ksep/K) or in the case when K = Fp we get Gal(Fp/Fp)
where Fp is the algebraic closure of Fp, since the separable closure agrees with the alge-
braic closure in this case.
3.2 Profinite Group Cohomology
Given a profinite group G, we may consider the category of all G-modules, abelian
groups with an action of G. The discrete abelian groups with a continuous action form a
full subcategory which is an abelian category. These type of G-modules are called discrete
G-modules and we will follow with a formal definition. In words, a discrete G-module, M,
is one such that the stabilizer of each element in M is an open subgroup of G.
Definition 3.2.1. A discrete G-module, M, is a G-module with the condition
M =⋃
MH
where the union is taken over all open subgroups H of G. M has the discrete topology and
the action of G on M is continuous.
Given M as above we define a complex C∗(G,M) in the following way. Let Cn(G,M)
denote the set of continuous maps from an n-fold cartesian product Gn to M. Define a
15
coboundary map d : Cn −→ Cn+1 by the formula
(d f )(g1, .., gn+1) = g1 f (g2, ..., gn+1)
= +∑i=n
i=1(−1)i f (g1, ..., gigi+1, ..., gn+1)
= +(−1)n+1 f (g1, ..., gn).
The cohomology groups of this complex are Hic(G,M) [11]. Alternatively, these cohomol-
ogy groups can be defined as the right derived functors of the functor MG, a fact which
will be used later. It will also be useful to understand how limits are incorporated into the
definition of profinite group cohomology.
Proposition 3.2.2. Let G be a profinite group and {Aα} be G-modules such that lim−→
Ai = A
then
Hq(G, A) = lim−→
Hq(G, Aα)
Proof. This is a consequence of Serre’s Prop. 8 in Sec. 2.2 of [11]. �
Proposition 3.2.3. Let A be a discrete G-module then
Hq(G, A) = lim−→
Hq(G/H, AH) for all q ≥ 0 (3.2.1)
where the limit is taken over all open normal subgroups H of G.
Proof. This is another consequence of Serre’s Prop. 8 in Sec. 2.2 of [11]. �
3.3 Computations in Continuous Galois Cohomology
We now apply profinite group cohomology to the case of G = Gal(Fp/Fp). Let Z be
the completion of Z for the topology of subgroups of finite index. Then Z � lim←−Z/nZ and
16
is therefore a profinite group. Since, as defined earlier, Gn � Z/nZ we get that
Gal(Fp/Fp) � lim←−
Gn � Z.
We also define Hn = Gal(Fp/Fpn) � nZ and we get from the isomorphism Gn � G/Hn (See
Prop. 2.1.2), that G = lim←−
G/Hn. We note that every open subgroup of G is of the form
Hn for some n. Given A a G-module, G has a single topological generator which is the
Frobenius automorphism on A (F(a) = ap where p is the characteristic of the base field).
A a discrete G-module necessitates that for a ∈ A there is a positive integer n such that
Fn(a) = a. This gives that A =⋃
AHn , where the union ranges over all n, as required. We
can now define the following
Definition 3.3.1. Let A be a discrete G-module, where G = Gal(Fp/Fp) and Hn = Gal(Fp/Fpn).
The cohomology of G with coefficients in A is
Hq(G, A) = lim−→
Hq(G/Hn, AHn) (3.3.2)
for all q ≥ 0.
In general, we have some useful propositions of Serre for low degree cohomology.
First, we note that for G any profinite group with an action on A degree 0 cohomology is
equal to the elements in A fixed by the action:
H0(G, A) = AG.
We now consider specifically G � Z � Gal(Fp/Fp). Let Nn be the map 1 + F + ... + Fn−1
and let
A′ = {a ∈ A|Nn(a) = 0 for some n > 0}.
17
Given that G acts on A by F, we have the following:
Proposition 3.3.2.
H1(G, A) � A′/(F − 1)A.
Proof. See [10]. �
We may also understand the cohomology in degree 2 for certain coefficient mod-
ules.
Proposition 3.3.3. Let A be a torsion group, then
H2(G, A) = 0.
Proof. Suppose A is finite then from Prop. 2.3.2 we have,
H2(Gn, AHn) = AG/NnAHn
where Nn is defined as above. We can show that the homomorphism
AG/NnAHn −→ AG/NnmAHnm
is just multiplication by m. We have that for a ∈ AG/NnAHn , F(a) = a and
(1 + F + ... + Fn−1)(a) = 0
so na = 0 and for the same reasons nma = 0 in AG/NnmAnmG. Thus, a must go to ma making
the map multiplication by m. If m is a multiple of the order of A the homomorphism is 0
and since this homomorphism is a part of the directed system of equation 3.3.2, which gives
us the definition of H2(G, A), then the claim is proved for A finite.
18
If A is torsion then A = lim−→
Aα for Aα finite. Thus, we have proven our claim by
Prop. 3.2.2.
H2(G, A) � lim−→
H2(G, Aα) � 0.
�
We now want to consider higher degree cohomology. From the following proposi-
tion, we get that cohomology vanishes in degree 3 and higher.
Proposition 3.3.4. Let A be a discrete G-module; then for each q ≥ 3
Hq(G, A) � 0.
Proof. We make use of computations of the cohomology of a cyclic group and determine
the homomorphisms from
Hq(Gn, AHn) −→ Hq(Gnm, AHnm).
We have the following due to Prop. 2.3.2,
Hq(Gn, AHn) �
AG/NnAHn for n = 0 mod 2
Nn AHn/D∗AHn for n = 1 mod 2.
When q=2 the map is multiplication by m as shown in the previous proof. Let q=3, then
we want to know the map induced in cohomology
Nn AHn/D∗AHn →Nnm AHnm/D∗AHnm .
If a ∈Nn AHn/D∗AHn , then (1 + ... + Fn−1)(a) = 0 and (F − 1)(a) = 0 so F(a) − a = 0, which
means F(a) = a, so na = 0. In Nnm AHnm/D∗AHnm , (1 + ...+ Fnm−1)(a) = 0 and (F − 1)(a) = 0.
19
Thus, the map after taking the quotient sends na to nma so before before taking quotients
the map must be multiplication by m. Thus after quotienting the map is multiplication by
m modulo the given cosets. We see here also that Nn AHn/D∗AHn is torsion, so when we go
far enough out in the limit the map will be multiplication by the order of this group and the
rest of the maps will be 0. This means that H3(G, A) = 0 for any G-module A. We proceed
with a shifting argument. Given an injective resolution of A,
0 −→ A −→ I0 −→ I1 −→ I2 −→ ...
we may truncate to get
0 −→ Af−→ I0 −→ coker( f ) −→ 0.
We then consider a part of the long exact sequence we get from this exact sequence:
... −→ H3c (G, coker( f )) −→ H4
c (G, A) −→ H4c (G, I0) −→ ...
We know H3c (G, coker( f )) � 0 since H3
c (G,M) � 0 for any discrete G-module M. Also,
H4c (G, I0) � 0 since I0 is injective. Thus, H4
c (G, A) � 0. Since this is true for any discrete
G-module A, we can use an inductive argument to show that for all q ≥ 3 Hqc (G, A) � 0.
Therefore, we have proven our desired result. �
We now compute the relevant cohomology for algebraic K-theory of finite fields.
Let A = µ⊗t` where µ` are the `-th roots of unity in Fp and where we think of A as a
G-module with a different action depending on the tensor power t. If t = 0 then G acts
trivially on A. If t = 1 then G acts on A by the Frobenius map F(a) = ap. If t > 1 then G
acts on A by F t. We then notice that the `-th roots of unity in Fp are isomorphic to Z/`Z as
20
abelian groups. We use the propositions above to compute the following.
Proposition 3.3.5. Let A = µ⊗t` be a G-module with the action described above. Then
H0(G, A) �
{a ∈ A|F t(a) = a} if t > 0
A if t = 0
�
Z/ gcd(`, pt − 1)Z if t > 0
Z/`Z if t = 0
H1(G, A) � {a ∈ A|(F t − 1)(a) = 0}
� Z/ gcd(`, pt − 1)Z
H s(G, A) � 0 if s > 1
Proof. Let A = µ⊗t` , first we may compute H0(G, A) easily by the fact that in general
H0(G, A) = AG, where AG are the elements fixed by the action of G and since the action is
trivial when t = 0 we get just A � Z/`Z. When t > 0 we get the elements in A such that
F t(a) = a which is isomorphic to the elements a ∈ Z/`Z such that (pt − 1)a = 0 so it is
isomorphic to Z/ gcd(`, pt − 1)Z. Thus, we have computed degree 0 cohomology.
For degree 1 cohomology, we need to use Prop. 3.3.2. First, we claim that A’=A
since in our case A is torsion. In general, let A be a topological G module. Then denote A f
the torsion subgroup of A. If a ∈ A f then na = 0 for some n. Also, Fm(a) = a for some m
since a ∈ A by definition of the Frobenius map. It follows that
(1 + F + ... + Fmn−1)a = (1 + F + .. + Fm−1 + Fm + Fm+1... + Fmn−1)(a)
= n(1 + F + ... + Fm−1)(a)
= na′ for some a′ ∈ A
= 0
which shows that a ∈ A′. Thus, the torsion elements of A are contained in A′; i.e. if A is
21
torsion then A ⊂ A′. Since A′ ⊂ A by definition we proved the claim. We now interpret the
result of Prop. 3.3.2:
H1(G, A) � A/(F t − 1)A.
In the proof of Prop. 3.3.2, we are using that G acts on A by the Frobenius map, so if G
acts on A by F t then we get the above definition. Thus,
H1(G, A) � (Z/`Z)/((F t − 1)Z/`Z)
We note that for a ∈ µ`, F t(a) − a = 0, so apt−1 = 0. Thus, for a ∈ Z/`Z, (pt − 1)a = 0, and
we get the following
H1(G, A) � Z/ gcd(`, pt − 1)Z
proving our claim.
We now need to show H s(G, A) = 0 for s > 1. From Prop. 3.3.3 we know that since
A is torsion,
H2(G, A) � 0.
Then we note that for any G-module A we have from Prop. 3.3.4
H s(G, A) � 0 for s ≥ 3.
Therefore, we have computed all the cohomology of G = Gal(Fp/Fp) with coeffictients µ⊗t`
as required. �
22
Chapter 4
Algebraic K-theory of finite fields
This computation of Algebraic K-theory for finite fields involves continuous galois
cohomology, etale cohomology and motivic cohomology, although it suffices to compute
continuous galois cohomology and then pass to etale or motivic cohomology where the
results will be equivalent in this special case. The ability to identify continuous galois co-
homology to etale cohomology or etale cohomology to motivic cohomology is nontrivial
and it shall be the first topic of discussion. Once we have a computation of etale or mo-
tivic cohomology, we use a spectral sequence where the E2 page can be identified with
etale or motivic cohomology with appropriate bi-grading. The reason for using either mo-
tivic or etale category in this setting is that we have two versions of an Atiyah-Hirzebruch
type spectral sequence to work with: one starting at motivic cohomology and one start-
ing at etale cohomology, each converging to similar versions of algebraic K-theory. The
computation proceeds as follows; (1) we identify continuous galois cohomology with etale
cohomology, (2) we identify etale cohomology with motivic cohomology, (3) we describe
the two versions of the spectral sequence we have, (4) we compute algebraic K-theory of
finite fields, and (5) we compare to results of Quillen.
23
4.1 Continuous Galois Cohomology to Etale Cohomology
For our purposes, we need enough etale cohomology to show an equivalence with
continuous galois cohomology. If we let G = Gal(Fp/Fp), then consider the category of
discrete G-modules, G-mod. It can be shown that there is an equivalence of categories
between G-mod and the category of sheaves of abelian groups over the etale site, Xet,
of X=spec(Fp). We denote this category S(Xet). We begin by defining some terms used
already to clarify the necessary propositions.
4.1.1 Definitions: basic algebraic geometry
Our main sources for these sections are Milne’s book on etale cohomology [7] and
Hartshorne’s Algebraic Geometry [2]. These definitions are meant to help the inexperi-
enced reader get a brief introduction to the material and it should be skipped by any reader
with previous knowledge of algebraic geometry on the level of sheaves and schemes. It
should be acknowledged that at the time of writing the author considers himself in the
former category.
Definition 4.1.1. A presheaf P is a contravariant functor from Top(X), the category of open
subsets of X with morphisms given by inclusions, to a category C , usually the category of
sets, abelian groups or commutative rings.
Definition 4.1.2. A sheaf F is a presheaf with the additional properties:
1. Given Ui a covering of U and s, t ∈ F (U) such that s|Ui = t|Ui for all i, then s = t, and
2. if for each i there exists a section si of F such that for Ui, U j and si|Ui∩U j = s j|Ui∩U j
then there exists a section s ∈ F (U) with s|Ui = si for each i.
24
Definition 4.1.3. The stalk of a sheaf F gives local information about the sheaf at a point.
It is given by the limit taken over all U such that the point x ∈ U
Fx = lim−→
F (U)
Definition 4.1.4. A ringed space is a topological space X, along with a structure sheaf of
rings OX. A ringed space is called a locally ringed space if each stalk of the structure sheaf
is a ring with a unique maximal ideal.
Definition 4.1.5. An affine scheme X is a locally ringed space isomorphic to spec(R) for
some commutative ring R.
Definition 4.1.6. A scheme X is a locally ringed space along with a covering of open sets
Ui such that restricting to the structure sheaf OX is an affine scheme.
Remark. This is a similar notion to a covering of a manifold with coordinate maps.
4.1.2 Definitions needed for etale cohomology
We have defined sheaves and schemes. We now want to interpret this in the etale
setting. To do this we need morphisms of schemes.
Definition 4.1.7. A morphism of schemes from X to Y is given by a pair (f, φ) where f is a
continuous map f : X −→ Y and φ : OX −→ f∗(OY) where f∗(OY) is the direct image of the
structure sheaf of X.
Definition 4.1.8. We say a morphism of schemes is etale if it is flat and unramified (this
also implies locally of finite type).
Definition 4.1.9. A morphism of schemes f : Y −→ X is affine if for any affine subset U
of X, then f −1(U) is affine in Y. If, for all U with this property, Γ( f −1(U),OY) is a finite
Γ(U,OX)-algebra, then f is finite.
25
For definitions of flat and unramified see [7]. It is enough to know that the Zariski
topology is too coarse for our purposes and by using etale morphisms as structure mor-
phisms for a scheme, we may get more interesting cohomology. For example, in the case
here where we look at spec(Fp) which is just a point in the Zariski topology. When we talk
about the etale topology, we mean in the sense of a Grothendieck topology on a category.
We define a topology on a subcategory of schemes over a base scheme. We call a category
along with a Grothendieck topology a site.
Definition 4.1.10. We define Xet to be the small local etale site of X. We fix the base scheme
X and consider the full subcategory of schemes over X with etale structure morphisms,
which we will denote et/X. This subcategory is given the etale topology in the following
way. We define an etale covering of an object Y on et/X as a family of etale morphisms
gi : Ui → Y i ∈ I with the property that Y =⋃
gi(Ui), for Ui ∈ et/X. The class of all such
coverings gi of all such Y forms the etale topology. The category along with this topology
is what we refer to as the small local etale site.
For the following proposition, we will also use the notion of a geometric point of a
space X with the etale topology. One difference between the etale topology and the Zariski
topology for example, is that usually when we consider sheaves over a point, we mean
just the set or abelian group assigned to that point, however in the etale setting, sheaves
over a point will only give a single set or abelian group if that point is spec(ksep) for ksep
a separably closed field. Otherwise, this will not be the case. For X=spec(k) where k is
not the separable closure, we denote the geometric point x = spec(ksep). In our example,
we are considering schemes over spec(Fp) so the geometric point will be x = spec(Fp). By
contravariance of the functor spec(-), we get a map ux : x −→ X induced by the inclusion
of Fp in Fp.
26
4.1.3 Equivalence of categories and cohomology theories.
We return to our goal of showing an equivalence of cohomology theories. It begins
by showing an equivalence of categories between G-mod and S(Xet) as laid out in the
introduction to this section.
Proposition 4.1.11. Let G = Gal(Fp/Fp) and X=spec(Fp). There is an equivalence of
categories between G-mod, the category of discrete G-modules and S(Xet), the category of
sheaves of abelian groups on the small local etale site.
Proof. We design functors from G-mod to S(Xet) and from S(Xet) to G-mod which give us
an equivalence of categories. First, we note that by fixing a separable closure of Fp (in this
case we have only one separable closure Fp) we are fixing a geometric point x → X and
we can write G as the fundamental group of X with respect to that geometric point G =
π1(X, x), in the sense of [7]. By contravariance, G acts on Fp on the left and spec(Fp) = x
on the right. Given a presheaf P on the site Xet, let
MP = lim−→
P(spec(Fpn))
for every finite separable extension of Fp. We inherit a left action of G on P(spec(Fpn))) from
the usual action of G on Fpn . Thus, we get an action of G on MP. It is clear that MP = ∪MHnP
where the union is taken over all subgroups Hn of G. This makes MP a discrete G-module.
On the other hand, given a discrete G-module M, we can construct a presheaf FM
with the following properties
1. FM(Fpn) = MHn , if Hn = Gal(Fp/Fpn)
2. FM(∏Fpi) =
∐FM(Fpi).
We do this in the following way. Let FM(U) be identified with G-module homomorphisms
27
from a functor F evaluated at U to the G-module M. We have
FM(U) = HomG(F (U),M)
where F is a functor from the category of X-schemes finite and etale over X,
FEt / X −→ G- sets
defined by F (U) = HomX(x,U). The category et/ X contains morphisms with an extra
property of finiteness as in Definition 4.9, and we call the category of such morphisms
FEt / X. G- sets is the category of finite sets with a continuous left action of G. The Hom
set is the set of X-scheme maps from the geometric point x into U a scheme over X. We
see then that F (Fpn) = HomX(spec(F p), spec(Fpn)) � G/Hn and F (∏
Fpi) =∏
F (Fpi).
Thus, defining FM gives us the presheaf we wish to find . It takes the work of another
lemma to prove that this is a sheaf, which can be found in [7]. We then have shown that we
can identify any G-module M with a sheaf in S(Xet) . To get an equivalence of categories
we need a little more work. We can see that given a morphism of discrete G-modules
f : M → M′ there is clearly an induced morphism on sheaves FM −→ FM′ . Given a
morphism on sheaves φ : F −→ F ′ we let Fpn = FHn where Hn = Gal(Fp/Fpn), then
φ(Fpn) : F (Fpn) −→ F ′(Fpn)
commutes with the action of G because φ is a functor. We can then define a G-homomorphism
by lim−→
φ(Fpn) from MF −→ MF ′ . We also can see that HomG(M,M′) −→ Hom(F ,F ′) is
an isomorphism and the map F −→ FMF is an isomorphism. Thus, we have shown that the
categories are equivalent. �
28
This equivalence of categories leads to an equivalence of cohomology theories of
etale cohomology of spec(Fp) and galois cohomology of Gal(Fp/Fp) with appropriate co-
efficients. In order to make this claim we must recall the definition of continuous galois
cohomology as right derived functors of the functor AG where A is the coefficient module.
We define a similar notion of etale cohomology in terms of right derived functors.
Definition 4.1.12. Let Γ(X,−) : S(Xet) −→ Ab be the functor defined by Γ(X,F ) = F (X).
This is a left exact functor and we can therefore define etale cohomology in terms of the
right derived functors Ri of Γ(X,−):
Hiet(X,−) � RiΓ(X,−).
Remark. Note that for definition 4.11 we need S(Xet) to have enough injectives. A proof of
this fact can be found in Milne [7].
Proposition 4.1.13. Let Fp be a finite field, Fp be the algebraic closure of Fp and G=Gal(Fp/Fp).
Let µ⊗tl be the group of l-th roots of unity in Fp with the G acting by F t. In etale cohomology
we see this as the sheaf of abelian groups over spec(Fp), which sends spec(Fpn) to the `-th
roots of unity in Fpn . Then
Hnc (G, µ⊗t
l ) � Hnet(Fp, µ
⊗tl ). (4.1.1)
Note that in etale cohomology we use Fp as shorthand for spec(Fp).
Proof. By Definition 4.12, we have,
RiΓ(X,−) = Hiet(X,−)
for the etale cohomology of X. From Proposition 4.11 we see that
Γ(X,F ) = F (X) = MG
29
so for X=spec(Fp), F = µ⊗t` , and M = µ
⊗t(Fp)` we have that
Hiet(X,F ) � RiΓ(X,F )
� RiMG
� Hic(G,M)
.
Thus, we have our desired result. �
4.2 Norm Residue Theorem
The following account shows a correspondence between etale cohomology and mo-
tivic cohomology used in computations of a spectral sequence which converges to algebraic
K-theory. The theorem of primary importance here is the Norm Residue Theorem. Let
Hnmot(k,Z/`(i)) be motivic cohomology defined to be the cohomology of the chain complex
Z/`(i) of Nisnevich sheaves. For more information on motivic cohomology and Nisnevich
sheaves see [5]. Many of the results in this section are not understood by the author in
their full generality. They are used here as a way to pass from galois cohomology where
computations are made, to motivic cohomology, which is the input for Weibel’s version of
the spectral sequence used in future computations.
Theorem 4.2.1. Norm Residue Theorem (Rost-Voevodsky) Let k be a field containing
1/`. The natural map then induces isomorphisms
Hnmot(k,Z/`(i)) �
{Hn
et(k, µ⊗i` ) n ≤ i
0 n > i (4.2.2)
If X is a smooth scheme over the base field k there is a the natural map
Hnmot(X,Z/`(i))→ Hn
et(X, µ⊗i` )
is an isomorphism for n ≤ i.
30
This form of the theorem is from Weibel’s book [16]. This gives us a correspon-
dence between motivic cohomology and etale cohomology. The coefficient change makes
sense due to a lemma in a paper by Voevodsky [14],
Lemma 4.2.2. Let k be a field and ` prime to the char(k), then there is a quasi isomorphism
in the etale topology on the category of smooth schemes over k
Z/`Z(i)et � µ⊗i` .
Thus, one makes the movement in the following way:
Hnmot(k,Z/`(i)) � Hn
et(k,Z/`(i)et) � Hnet(k, µ
⊗i` ).
We interpret these results for the case of finite fields in the following corollary.
Corollary 4.2.3. Let Fp be a finite field of characteristic p, X=spec(Fp), and let ` be a
prime with ` , p (Note that 1/` ∈ Fp). Then the natural map induces an isomorphism
Hnmot(X,Z/`(i)) �
Hn
et(X, µ⊗i` ) n ≤ i
0 n > i(4.2.3)
This gives us the relevant isomorphism for our computation. In Weibel’s version of
the spectral sequence, we need this intermediary step to get to motivic cohomology which
is the input in the spectral sequence. We also need the intermediary step to get from etale
cohomology to galois cohomology, from the previous section. Ultimately, we will use a
version of the spectral sequence which takes etale cohomology as input. The fact that the
spectral sequences used in the next section compute algebraic K-theory depends on the
validity of the norm residue theorem.
31
4.3 Algebraic K-theory Spectral Sequence
We have two different versions of an Atiyah-Hirzebruch type spectral sequence.
One version comes from Weibel, which he calls the “motivic to algebraic K-theory” spec-
tral sequence. His version computes algebraic K-theory with coefficients outright which is
an advantage, but his differentials are not specified explicitly, which is a disadvantage. We
get from Weibel [16] the following:
Theorem 4.3.1. Let A be any coefficient group and X be a smooth scheme over a field k.
Then there is a spectral sequence from motivic cohomology to algebraic K-theory with E2
term
E s,t2 = H s−t
mot(X, A(−t)⇒ K−s−t(X; A). (4.3.4)
If X=spec(k) and A = Z/`Z, where 1/` ∈ k then the E2 terms are just the etale cohomology
groups of k, which lie only in the octant where s ≤ t ≤ 0.
This is enough to see what the E2 page would be, but without knowing what the
differentials are we cannot be sure what will happen on the Er page for r > 2. From
looking at another version of this type of spectral sequence due to Thomason [13] we can
see that the differentials should be trivial in our specific case.
Theorem 4.3.2. Let ` , p be prime, and let X be a smooth scheme satisfying some other
mild hypotheses containing 1/` (X=spec(k) for k a field satisfies the necessary hypotheses).
Let β be the Bott element. The graded ring K/`∗(x) may be localized by inverting this β. We
can then compute K/`∗(X)[β−1] in terms of etale cohomology of X:
E s,t2 =
H s
et(X;Z/`(i)), t = 2i
0, t odd⇒ K/`t−s(X)[β−1]. (4.3.5)
What we lose here is that we are now computing mod ` algebraic K-theory with an
32
inverted Bott element instead of outright algebraic K-theory, but this will be enough for our
purposes due to Theorem 4.4.2 and Theorem 4.5.2. What we gain here is that Thomason
gives the differentials in the spectral sequence:
dr : E s,tr −→ E s+r, t+r−1
r
in his paper [13]. We can then set up the spectral sequence in the same way one would set
up the Adams spectral sequence with s on the vertical axis and t-s on the horizontal axis.
In the next section, we use the information from previous sections to input in our spectral
sequence and compute algebraic K-theory of finite fields.
4.4 Spectral Sequence Computation
We begin by interpreting Thomason’s spectral sequence in the case of finite fields
where etale cohomology can be identified with continuous galois cohomology.
Corollary 4.4.1. Let G = Gal(Fp/Fp) and let µ` be the `-th roots of unity in Fp. Then the
following is a spectral sequence from continuous galois cohomology, which converges to
mod ` algebraic K-theory with an inverted Bott element
E s,t2 =
H s
c(G; µ⊗i` ), t = 2i
0, t odd, s > i⇒ K/`t−s(Fp)[β−1]. (4.4.6)
We know that we only have degree 0 and degree 1 continuous galois cohomology
with these coefficients due to Prop. 3.3.3 and Prop. 3.3.4, which means that the spectral
sequence is only nontrivial with s in degree 0 and 1 and with t in even degrees. We have the
following which tells us that in most degrees our answer will agree with algebraic K-theory
without the inverted Bott element.
33
Theorem 4.4.2. Let Fp be a finite field and K/`n(Fp)[β−1] be algebraic K-theory with an
inverted Bott element; then
K/`n(Fp)[β−1] = Kn(Fp,Z/`Z)
for n > 0.
Proof. To show that algebraic K-theory of finite fields with an inverted Bott element is
equivalent to algebraic K-theory without the inverted Bott element, we need to see where
the Bott element comes from. Due to Weibel, we know the Bott element lives in degree 2 of
algebraic K-theory of finite fields and multiplication by the Bott element gives an isomor-
phism to higher even degrees [16]. Similarly, the Bott element gives an isomorphism from
degree 1 algebraic K-theory to higher odd algebraic K-theory. Thus, inverting the Bott
element does not change the K-theory groups since there is no Bott torsion in algebraic
K-theory above degree 0. �
We now have all the information necessary to make our computation.
Theorem 4.4.3. Let X=spec(Fp), ` , p a prime, then we compute the folllowing,
K/`t−s(X)[β−1] �
Z/` if t − s = 0
Z/ gcd(`, pi − 1) if t − s = 2i − 1
Z/ gcd(`, pi − 1) if t − s = 2i
Proof. Using Thomason’s version of the spectral sequence and setting up the axis with t-s
on the horizontal axis and s on the vertical axis, we can then input our answers from the
34
previous section into the nontrivial positions. Let us recall these calculations,
H0c (G, µ⊗i
` ) �
Z/`Z if i = 0
Z/ gcd(`, pi − 1)Z if i > 0
H1c (G, µ⊗i
` ) � Z/ gcd(`, pi − 1)Z
From Thomason we have
E s,t2 �
H s
c(G, µ⊗i` ) if t = 2i
0 if t = 2i − 1⇒ K/`t−s(Fp)[β−1]
Thus, the following table will help us input the correct information. Let i ≥ 1.
s t t − s Group
0 0 0 Z/`
0 2i − 1 2i − 1 0
0 2i 2i Z/gcd(`, pi − 1)
1 1 0 0
1 2i 2i − 1 Z/gcd(`, pi − 1)
1 2i − 1 2i 0
We get the following E2 page with d2 differentials given by the arrows. The picture is
shown with s on the vertical axis and t − s on the horizontal axis.
35
0 2 4 6
0
2
4
Z/`
Z/ gcd(`, p − 1)
Z/ gcd(`, p − 1)
Z/ gcd(`, p2 − 1)
Z/ gcd(`, p2 − 1)
Z/ gcd(`, p3 − 1)
Z/ gcd(`, p3 − 1)
We see that there is no room for nontrivial differentials, thus each group will persist
to the next page. Therefore, E2 � E3 and since no dr differentials for r > 2 will be nontrivial
we get E2 � E∞. Thus, we get our computation of K/`t−s(Fp)[β−1],
K/`t−s(Fp)[β−1] �
Z/` if t − s = 0
Z/ gcd(`, pi − 1) if t − s = 2i − 1
Z/ gcd(`, pi − 1) if t − s = 2i
Since ` is prime,
gcd(`, pi − 1) =
` if `|pi − 1
1 otherwise.
Thus, wherever ` does not divide pi − 1 for some i the mod ` algebraic K-theory vanishes.
36
�
We notice that there are no nontrivial differentials anytime the cohomological di-
mension n ≤ 1. In this situation we need only consider cohomological dimension of a
profinite group.
Definition 4.4.4. The `-cohomological dimension of a profinite group G is less than or
equal to n, written cd(G)≤n if and only if Hr(G, A) = 0 for all r > n and all `-torsion
G-modules A.
Since we cannot always identify etale cohomology with profinite group cohomol-
ogy, we need to consider etale cohomological dimension as well. The relevant cohomo-
logical dimension here is `-cohomological dimension. We say a sheaf of abelian groups F
is `-torsion if for all quasi-compact U F (U) is `-torsion. Clearly, the coefficients we are
working with have this property.
Definition 4.4.5. The `-cohomological dimension of a site, (C/X)et is the smallest integer
n (or∞) such that Hi(Xet,F ) = 0 for all i > n and all `-torsion sheaves F .
Using this definition of cohomological dimension we get an easy corollary to The-
orem 4.3.2:
Corollary 4.4.6. Whenever Hnet(X,Z`(i)) has cohomological dimension n ≤ 2, the differen-
tials will be trivial in all degrees in the spectral sequence of Theorem 4.3.2, and E2 � E∞.
4.5 Comparison to results of Quillen
Computing higher algebraic K-theory by definition requires heavier machinery,
therefore we prefer to use the results cited in this paper for the computation. However, the
only way to compute higher algebraic K-theory groups originally was by definition. The
37
work of computing and defining higher K-theory comes from Quillen [9], [8]. Though
the work in these papers is not the focus of this thesis, it is interesting to see the results
that Quillen gets in his computation. One difference in our computation is that we are com-
puting mod ` K-theory. It is for this reason that we see Z/`Z in our result depending on a
relation to the characteristic of the finite field we consider. In Quillen’s computation in [8],
he arrives at
Theorem 4.5.1. (Quillen) If k is a finite field with p elements then
K0(k) � Z
K2i(k) � 0
K2i−1(k) � Z/(pi − 1)Z.
for i ≥ 1.
We use two theorems to show the relationship between this calculation and our
calculation. One stated earlier, Theorem 4.4.2, gives us algebraic K-theory without the
inverted Bott element above degree 0. The second shows how to get from algebraic K-
theory with coefficients to algebraic K-theory without coefficients.
Theorem 4.5.2. (Universal Coefficient Theorem for algebraic K-theory)
The following is a short exact sequence
0→ Kn(k) ⊗ Z/` → Kn(k;Z/`Z)→ `Kn−1(k)→ 0
where `Kn−1(k) is the ` torsion elements of Kn−1(k) for all n ∈ N and ` and any ring k.
Our reference for this theorem is Weibel’s book on algebraic K-theory [16]. We
use this theorem to make the following computation which relates the results of Quillen to
the results found in this paper:
38
Proposition 4.5.3. Let ` , p be a prime. Then it follows from Quillen’s computation of
K∗(Fp) that
K∗(Fp;Z/`) �
Z/` if, `|pi − 1 and ∗ = 2i
Z/` if , `|pi − 1 and ∗ = 2i − 1
0 if ` does not divide pi − 1
,
for ∗ > 0.
Proof. We show this using the Universal Coefficient theorem which gives us
0→ Kn(Fp) ⊗ Z/` → Kn(Fp;Z/`)→ `Kn−1(Fp)→ 0.
For even n=2i+1, i ≥ 0 we get the following from Quillen’s computation,
0→ Z/(pi − 1)Z ⊗ Z/` → Kn(Fp;Z/`)→ 0→ 0.
Therefore, by exactness the map is an isomorphism,
0→ Z/ gcd(pi − 1, `)Z � Kn(Fp;Z/`)→ 0.
Note that Z/(pi − 1)⊗Z/` � Z/ gcd(pi − 1, `). Since we assume ` is prime and not equal to
p, we get
Z/ gcd(pi − 1, `) �
Z/` if `|pi − 1
0 otherwise
We also have for n=2i, i ≥ 1,
0→ 0→ Kn(Fp;Z/`)→ `Z/(pi − 1)→ 0
39
which gives us,
0→ Kn(Fp;Z/`) � Z/` → 0
if `|pi − 1 and Kn(Fp;Z/`) � 0 otherwise. This follows because the `-torsion elements in
Z/(pi − 1)Z will be a copy of Z/ gcd(`, pi − 1)Z, which as we said before Z/`Z if `|pi − 1
and 0 otherwise.
In sum,
Kn(Fp;Z/`) � K/`n(Fp) � Z/`Z
for `|pi − 1 where n = 2i or n = 2i − 1. �
This shows us that in computing the mod ` algebraic K-theory via continuous ga-
lois cohomology our claims are validated by previous work of Quillen. The result of this
paper therefore is not new, but the approach generalizes to other possible computations as
described in the next section.
40
Chapter 5
Future Research
The computation we present here is a natural first computation to do using the
methods outlined in this paper. Further computations could be made using knowledge of
etale cohomology. In this case, etale cohomology was easily computed via continuous
galois cohomology, but it would be interesting to consider other cases where we cannot
take this approach. Milne computes etale cohomology of curves over an algebraically
closed field with coefficients in µ`. The only augmentation to this result needed for the
spectral sequence is to consider the higher tensor power coefficients. Milne’s computation
in his lecture notes [6] is the following
Theorem 5.0.4. If X is a complete connected nonsingular curve over k with k algebraically
closed and ` a prime different from char(k),
H0(X, µ`) � µ`(k)H1(X, µ`) � (Z/`Z)2g
H2(X, µ`) � (Z/`ZH s(X, µ`) � 0
for s > 2, and g the genus of the curve X.
The next step would be to compute algebraic K-theory for these types of curves.
41
Another direction would be unveil the underpinnings of the spectral sequence itself
in order to understand cases where the differentials are nontrivial, or perhaps to find oper-
ations preserved by the differentials. It would also be interesting to understand Quillen’s
original methods and constructions better.
The methods used in this work were preferred because they introduced cohomol-
ogy computations and spectral sequence computations, which are important to work in
algebraic topology and algebraic geometry. It served as an introduction to different coho-
mology theories and different categories in which a mathematician may work and showed
how one may move between them.
42
Bibliography
[1] Kenneth S. Brown. Cohomology of groups, volume 87 of Graduate Texts in Mathe-
matics. Springer-Verlag, New York, 1994. Corrected reprint of the 1982 original.
[2] Robin Hartshorne. Algebraic geometry. Springer-Verlag, New York, 1977. Graduate
Texts in Mathematics, No. 52.
[3] Thomas W. Hungerford. Algebra, volume 73 of Graduate Texts in Mathematics.
Springer-Verlag, New York, 1980. Reprint of the 1974 original.
[4] Serge Lang. Topics in cohomology of groups, volume 1625 of Lecture Notes in Math-
ematics. Springer-Verlag, Berlin, 1996. Translated from the 1967 French original by
the author, Chapter X based on letters written by John Tate.
[5] Carlo Mazza, Vladimir Voevodsky, and Charles Weibel. Lecture notes on motivic
cohomology, volume 2 of Clay Mathematics Monographs. American Mathematical
Society, Providence, RI, 2006.
[6] James S. Milne. Lectures on etale cohomology. May 3, 2012.
[7] James S. Milne. Etale cohomology, volume 33 of Princeton Mathematical Series.
Princeton University Press, Princeton, N.J., 1980.
43
[8] Daniel Quillen. On the cohomology and K-theory of the general linear groups over a
finite field. Ann. of Math. (2), 96:552–586, 1972.
[9] Daniel Quillen. Higher algebraic K-theory: I [mr0338129]. In Cohomology of groups
and algebraic K-theory, volume 12 of Adv. Lect. Math. (ALM), pages 413–478. Int.
Press, Somerville, MA, 2010.
[10] Jean-Pierre Serre. Local fields, volume 67 of Graduate Texts in Mathematics.
Springer-Verlag, New York, 1979. Translated from the French by Marvin Jay Green-
berg.
[11] Jean-Pierre Serre. Galois cohomology. Springer-Verlag, Berlin, 1997. Translated
from the French by Patrick Ion and revised by the author.
[12] Stephen S. Shatz. Profinite groups, arithmetic, and geometry. Princeton University
Press, Princeton, N.J., 1972. Annals of Mathematics Studies, No. 67.
[13] R. W. Thomason. Algebraic K-theory and etale cohomology. Ann. Sci. Ecole Norm.
Sup. (4), 18(3):437–552, 1985.
[14] Vladimir Voevodsky. Bloch-Kato conjecture for Z/2-coefficients and algebraic
Morava K-theories. Preprint, June 17, 1995, K-theory Preprint Archives,
http://www.math.uiuc.edu/K-theory/0076/.
[15] Vladimir Voevodsky. The Milnor Conjecture. Preprint, December 20, 1996, K-theory
Preprint Archives, http://www.math.uiuc.edu/K-theory/0170/.
[16] Charles W. Weibel. Algebraic k-theory. Rutgers University.
44